CAP 4770 - Lecture 2: Examine data, attributes, types, data quality, and similarity measures

Vocabulary flashcards covering key terms from Lecture 2: data concepts, attribute types, data representations, data quality issues, preprocessing, and similarity measures.

Data

A collection of data objects and their attributes.

Object

An individual data item; also called a record, data point, case, sample, entity, or instance.

Attribute

A property or characteristic of an object; also known as a variable, field, characteristic, or feature.

Attribute values

Numbers or symbols assigned to an attribute; the same attribute can map to different values, and different attributes can map to the same value set.

Categorical (qualitative)

Attributes whose values are categories or labels (e.g., eye color, gender, zip codes).

Quantitative

Attributes whose values are numerical measurements (e.g., age, income, length).

Discrete attribute

An attribute with a finite or countably infinite set of values; often represented as integers (binary attributes are a special case).

Continuous attribute

An attribute with real-number values; in practice represented with finite precision (floating-point).

Nominal

Attribute with only distinct names and no inherent order (e.g., ID numbers, eye color).

Ordinal

Attribute that provides a meaningful order of objects (e.g., rankings, grades).

Interval

Differences between values are meaningful; no true zero (e.g., calendar dates, Celsius).

Ratio

Differences and ratios are meaningful and there is a true zero (e.g., age, length, weight).

Attribute transformation

Process of mapping an attribute’s values to a new set of values to enable analysis (e.g., normalization, log, exponentiation).

Normalization

Transforming values to a standard scale, often [0,1] or with zero mean and unit variance.

Data Matrix

A matrix where rows are objects and columns are attributes; used for numeric attributes; m×n dimensions.

Record Data

Data consisting of a collection of records, each with a fixed set of attributes.

Document Data

Each document is represented as a term vector; the value of a term component is its frequency in the document.

Transaction Data

A special form of record data where each record is a transaction—a set of items bought together.

Graph Data

Data represented as graphs consisting of nodes (entities) and edges (relationships).

Ordered Data

Data where the order of values matters (e.g., genomic sequences).

Spatio-Temporal Data

Data with both spatial and temporal components (space and time).

Data Quality

Quality of data; problems include noise, missing values, outliers, and duplicates.

Noise

Modification or distortion of original values (e.g., voice distortion, television snow).

Outliers

Data objects with characteristics that are markedly different from the rest of the data.

Missing Values

Absent information; causes include nonresponse or inapplicability; handling includes elimination, estimation, ignoring, or imputation.

Duplicate Data

Duplicate or near-duplicate objects; common issue when merging data from different sources.

Data Preprocessing

Preparing data for mining, including discretization, binarization, and attribute transformation.

Discretization

Transforming a continuous attribute into a discrete set of values or categories.

Binarization

Converting attribute values to binary (0/1) values.

Similarity

Numerical measure of how alike two data objects are; higher means more alike.

Dissimilarity

Numerical measure of how different two data objects are.

Euclidean Distance

Distance between two objects in continuous feature space; may require standardization if scales differ.

Standardization

Transforming features so they have zero mean and unit variance.

Binary Vectors

Data objects with binary attributes; similarity can be computed from M01, M10, M00, M11 counts.

Simple Matching Coefficient (SMC)

SMC = (M11 + M00) / total attributes; proportion of matches between two binary vectors.

Jaccard Coefficient

J = M11 / (M01 + M10 + M11); proportion of 1-1 matches among non-both-zero attributes. It is a measure of similarity for binary vectors, representing the proportion of attributes where both objects have a '1' value (true positives) relative to the total number of attributes where at least one object has a '1' value (true positives + false positives + false negatives). It specifically excludes cases where both objects have a '0' (M00).

Mij counts

M01, M10, M00, M11 are counts used to compute similarity/dissimilarity for binary data: M01 (number of attributes where object 1 is 0 and object 2 is 1), M10 (number of attributes where object 1 is 1 and object 2 is 0), M00 (number of attributes where both objects are 0), and M11 (number of attributes where both objects are 1).